Fatigue is one of the biggest safety and performance risks in mining operations and other heavy industries. Workers who are tired are more likely to make mistakes, have slower reactions, or suffer injuries on the job. Traditionally, mining fatigue management has relied on tools like wearable sleep trackers, cameras, or reactive measures that only spot fatigue after it starts.
But recent peer-reviewed research shows something important: you can accurately predict fatigue risk even when wearables aren’t available, using a structured survey-based model combined with machine learning and a proven fatigue science algorithm. This approach supports predictive fatigue risk management and gives safety leaders reliable insights into worker readiness without requiring devices that may be hard to deploy.
A peer-reviewed study analyzed survey data from 1,800+ mining workers across 10+ mine sites
The research shows that fatigue risk can be predicted accurately without wearable sleep data
A survey-based model was used to create sleep profiles
Multiple previous research reports show that feeding these sleep profiles into a validated fatigue science model (SAFTE™) can output accurate fatigue predictions
The resulting fatigue scores showed over 90% sensitivity, low error, and fewer false positives than some wearable-based approaches
This approach is especially useful for mining operations where wearables are difficult to deploy due to privacy concerns, union environments, connectivity limits, or inconsistent device use
Predictive fatigue risk management refers to systems that forecast when a worker might be at elevated risk of fatigue before the risk affects performance or safety. With effective predictive models, supervisors can plan safer schedules, adjust tasks, and reduce incidents tied to worker fatigue.
Traditional methods include:
Wearable devices that track sleep patterns, like the ReadiWatch
Cameras or sensors that react to visible fatigue
Self-reported sleep logs
Each of these has limitations in mining environments:
Wearables may face privacy concerns, compliance challenges, or may be forgotten
Cameras and sensors are reactive and may not work everywhere
Self-reports can be inconsistent
The new research explores whether a survey-based method can provide accurate sleep profiles, which can then be fed into fatigue prediction models, even without wearable data.
The research, titled “Non-Wearable Survey-Based Deep Learning Techniques for Measuring Fatigue – Predicting Mining Workers’ Readiness,” was conducted by:
Authors:
Umme Zakia, New York Institute of Technology & Fatigue Science
Luis Lopez, Fatigue Science
Paul Kenny, Fatigue Science
David McGrail, Fatigue Science
This study validates the use of a survey-based questionnaire to accurately predict a person’s sleep profile, including sleep quantity and quality. It does not model fatigue risk directly.
The conversion of sleep profiles into fatigue risk scores (ReadiScores) is supported by separate peer-reviewed research, including studies linking sleep to fatigue and cognitive effectiveness through biomathematical models such as SAFTE™. Readi’s approach is therefore grounded in multiple peer-reviewed sources, each validating a different part of the system.
What the Researchers Studied
The study analyzed structured survey data collected from more than 1,800 mining workers across 10+ mining sites worldwide. These workers included both day-shift and night-shift employees.
The goal was to test whether survey responses could be used to estimate sleep patterns well enough to support predictive fatigue risk modeling, even when no wearable sleep data was available.
Rather than asking vague questions like “Do you feel tired?”, the survey focused on specific, behavior-based information, including:
Sleep timing and duration
How long it takes to fall asleep
How often workers wake during sleep
Commute times
Sleep environment
Shift schedules and rotation
Lifestyle factors that affect sleep
In total, 50 survey questions were used to build a detailed picture of how workers sleep and recover between shifts.
Machine-learning models then used these survey responses to estimate a person’s sleep profile, including things like minutes resting, sleep efficiency, and number of awakenings during sleep.
These estimated sleep profiles were fed into a scientifically validated fatigue algorithm called the SAFTE™ model, which has been used in aviation, military, and industrial fatigue science. The SAFTE™ model calculates a ReadiScore, an hour-by-hour prediction of an individual’s fatigue risk.
By combining survey-based sleep estimates with machine learning and the SAFTE™ science, the study created a predictive fatigue model that does not require wearable sleep tracking.
This research is peer-reviewed, meaning other scientists evaluated the methods and results before publication, which strengthens its credibility.
Here are the key findings:
The survey-based model was able to identify fatigue risk with over 90% sensitivity when compared with wearable-based predictions. That means it matched the wearable-based model’s alerts most of the time.
The average difference (Mean Absolute Error) between survey-based scores and wearable-based scores was under 10 points on the fatigue scale. This is a small difference in practical terms.
The model produced fewer false alarms than some wearable-based comparisons. False positives can distract supervisors and reduce trust in the system.
Read the ROI of Readi whatepaper
The model worked well across different sites and over long periods of data collection, which shows it is robust enough for real mining operations.
Mining is a 24/7 operation with shift rotations, long hours, and high physical demands. Fatigue is linked to accidents, injuries, and reduced productivity.
A predictive model that works without wearables means:
Safety teams can forecast high-risk periods
Crew schedules can be optimized using data
Supervisors get ahead of fatigue rather than reacting after it shows up
Not all mining sites can deploy wearables due to:
Privacy concerns from workers
Union and labour sensitivities
Poor connectivity underground
Device compliance issues
A survey-based system avoids these hurdles while still giving accurate fatigue insight.
When systems do not rely on cameras or wearables, workers may be more likely to participate honestly, improving data quality and program adoption.
This survey-based approach does not need to replace all other tools. Instead, it can:
Serve as a standalone fatigue risk management system where wearables are not feasible
Act as a backup when wearable data is missing
Support hybrid programs that combine wearables and non-wearable inputs
The key point is that fatigue risk visibility does not disappear when wearables are unavailable.
Managing fatigue risk in mining and heavy industry is critical for safety, productivity, and worker health. This peer-reviewed study shows that a survey-based predictive model can deliver fatigue risk forecasts that match traditional wearable-based methods, while avoiding some practical barriers to implementation.
If your team is evaluating fatigue risk tools, it’s worth considering how survey-based predictive methods can fit into your fatigue risk management strategy and strengthen your safety outcomes.
Yes. This peer-reviewed study shows that fatigue risk can be predicted accurately without wearable monitoring. Using a structured questionnaire, machine-learning models estimate key sleep inputs and feed them into a validated fatigue science model. The resulting ReadiScores closely matched predictions based on wearable sleep data, with over 90% sensitivity.
Fatigue Science uses a survey-based approach to fatigue monitoring when wearables are not available. Workers complete structured questionnaires about sleep and work patterns, which are analyzed using machine learning. These estimates are then processed through the SAFTE™ fatigue model to generate ReadiScores, providing hour-by-hour fatigue risk predictions without requiring devices.
Yes. The research shows that survey-based fatigue monitoring can reliably support fatigue management programs. The survey-derived sleep profiles were strongly correlated with wearable-derived sleep data, resulting in low error rates and fewer false positives. This makes survey-based monitoring a viable option for operations where wearables are difficult to deploy.
Readi is a fatigue risk management system developed by Fatigue Science. It uses sleep science and predictive modeling to assess fatigue risk before it impacts safety or performance. Readi can operate with wearable data, survey-based data, or a combination of both, giving organizations flexibility in how they manage fatigue.
According to the peer-reviewed study, Readi’s survey-based fatigue predictions achieved over 90% sensitivity compared to wearable-based models, with low average error and fewer false alarms. This level of accuracy supports its use in real-world fatigue management and fatigue monitoring programs.